agentwire-project-config

Reference instructions for two AgentWire project configuration files: one for personal project settings and another for protected task-execution commands. They explain settings such as roles and worktrees, and how scheduled tasks are authored and promoted.

In plain words
What is it for?
Configuring AgentWire projects, managing roles and worktrees, defining scheduler hooks, and keeping live personal settings separate from protected execution rules.
Why use it?
Mixing personal settings with executable task commands can cause stale configurations to run or expose machine-specific details in a repository. Separating the files clarifies what agents may edit and what requires protected handling.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/dotdevdotdev/agentwire-dev/agentwire-project-config
Any agent
npx skills add dotdevdotdev/agentwire-dev --skill agentwire-project-config
Clone the repo
git clone --depth 1 https://github.com/dotdevdotdev/agentwire-dev

Made for: Claude Code, Codex.

Per session 105 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,451 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 2 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00105 $0.04451
Opus 5 $0.00053 $0.02226
Sonnet 5 $0.00021 $0.00890
Haiku 4.5 $0.00011 $0.00445

Measured yesterday against content hash 3c8084a7f7af, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade C, and why

agentwire-project-config scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

logs`) so widening is copy-paste. Hard blocks (`rm -rf`, `git push --force`) and

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

weather: "curl -s wttr.in/?format=3"
.claude/skills/agentwire-project-config/SKILL.md · 307 lines

How it starts

The opening of the file, as written. The whole thing — 307 lines — stays where its author put it; the contents beside it link to each section on GitHub.

.agentwire.yml Project Config

Each project can have a .agentwire.yml in its root directory. This configures posture, roles, voice, and parent for that project.

Split from task-execution config (#720). .agentwire.yml is now PURELY declarative — posture/roles/voice/parent/worktree, no execution vector — so it's agent-writable. Named tasks: (the code the scheduler actually runs via shell=True: pre/post/on_task_end/shell) live in a separate file, .agentwire.tasks.yml, which is protected control-plane. See Task Schema below for how to author it.

Gitignore it (don't commit it)

.agentwire.yml should be gitignored, not committed. Two reasons:

  1. A tracked copy breaks worktree dispatch subtly. Worktree-dispatched runs (scheduler tasks, worktree sessions) check out HEAD — if the file is tracked, uncommitted live edits are invisible to the run, which silently executes the stale committed version. The failure mode is confusing: the live file looks right, the run behaves wrong.
  2. It's personal/live config, not project code. Voice choices, email recipients, schedules, machine-specific roles — none of that belongs in a repo, especially a public one.

Worktree runs still get the file: projects.worktrees.copy_files (default [".env", ".agentwire.yml", ".agentwire.tasks.yml"]) copies gitignored configs from the main checkout into every new worktree as live files — the live file always wins, no drift.

AgentWire enforces this automatically: whenever it writes .agentwire.yml into a git repo (agentwire projects create, agentwire new --persist, portal project settings), it appends the file to the project's .gitignore unless it's already ignored — or already tracked. A tracked file is left alone: committing is a valid choice for teams that want shared, versioned task definitions, but they're opting into the HEAD-checkout behavior above — worktree runs use the committed version, never live edits. To switch an existing project to the recommended setup: git rm --cached .agentwire.yml, add it to .gitignore, commit. agentwire tasks promote gitignores .agentwire.tasks.yml (via a .agentwire.tasks*.yml glob, covering the staging draft too) the same way, on first promote.

Read the full file on GitHub · 307 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 307 lines · 105 tokens per session scan C 3c8084a7f7af

Subscribe to this mod's changes

agentwire-project-config is a skill published in the GitHub repository dotdevdotdev/agentwire-dev (21 stars, last pushed 17d ago), licensed Apache-2.0. It adds 105 tokens to every session and 4,451 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

luvus-module

Write a luvus module (an extension for luvus, mission control for your AI coding agents). Use when the user is building or debugging a luvus module: authoring luvus-module.toml, adding a sidebar dock, Luvus Bar widget, right-click action, event hook, module pane, module settings, or calling luvus back over its UHP.

RizRiyz/luvus · 79 tokens

luvus

Control Luvus through its local CLI and UHP. Use only for a line beginning with =target message, an explicit request naming Luvus, a request to delegate to a named live Luvus agent or pane, or an explicit Luvus operation involving sessions, workspaces, tabs, panes, agents, files, Git, DIFF, worktrees, tasks, leases…

RizRiyz/luvus · 177 tokens

tmux-via-mcp

Use the tmux MCP tools to create sessions, shape layouts, run tracked commands, and automate interactive terminals when a real TTY or parallel panes are required.

bnomei/tmux-mcp · 39 tokens

tmux-buffer-explorer

Explore large tmux buffers via search and bounded slices. Use when buffer data is too large to load at once or needs incremental inspection.

bnomei/tmux-mcp · 33 tokens

not-happy-jan

Adjust Not-Happy-Jan feedback settings by running the nhj CLI when the user wants to tune voices, dials, muting, censoring, tests, or hold music.

guruswami-ai/not-happy-jan · 42 tokens

release

Use when the user asks to release unitmux, bump the version, publish a GitHub release, update the Homebrew cask, check release status, or says リリース/release/バージョン上げて.

yugo-ibuki/unitmux · 49 tokens